scispace - formally typeset
Journal ArticleDOI: 10.1109/TAI.2021.3064003

Effectiveness of Deep Learning on Serial Fusion Based Biometric Systems

04 Mar 2021-Vol. 2, Iss: 1, pp 28-41
Abstract: We develop a framework for multibiometric systems, which combines a deep learning technique with the serial fusion method. Deep learning techniques have been used in unimodal and parallel fusion-based multimodal biometric systems in the past few years. While deep learning techniques have been successful in improving the authentication accuracy, a biometric system is still challenged by two issues: 1) a unimodal system suffers from environmental interference, spoofing attacks, and nonuniversality, and 2) a parallel fusion-based multimodal system suffers from user inconvenience as it requires the user to provide multiple biometrics, which in turn takes longer verification times. A serial fusion method can improve user convenience in a multibiometric system by requiring a user to submit only a subset of the available biometrics. To our knowledge, the effectiveness of using a deep learning technique with a serial fusion method in multibiometric systems is still underexplored. In this article, we close this research gap. We develop a three-stage multibiometric system using a user's fingerprint, palm, and face and test three serial fusion methods with a Siamese neural network. Our experiments achieve an AUC of 0.9996, where the genuine users require only 1.56 biometrics (instead of all 3) on an average. Impact statement— We work on enhancing the user convenience and reducing the verification error in a multibiometric system. An improved multibiometric system can help law enforcement, homeland security, defense, and our daily lives by providing better access control. With the advent of deep learning technologies, the accuracy of multibiometric systems have been improved significantly; however, its applicability is still in question because of long verification times required by parallel fusion in a multibiometric system. Our proposed multibiometric framework alleviates this user inconvenience issue by utilizing a serial fusion strategy in decision making and improves accuracy by leveraging deep learning technology in feature extraction and score generation.

... read more

Topics: Biometrics (51%)
Citations
  More

6 results found


Open accessJournal ArticleDOI: 10.1109/COMST.2021.3108618
Van-Linh Nguyen1, Po-Ching Lin1, Bo-Chao Cheng1, Ren-Hung Hwang1  +1 moreInstitutions (1)
Abstract: Sixth-generation (6G) mobile networks will have to cope with diverse threats on a space-air-ground integrated network environment, novel technologies, and an accessible user information explosion. However, for now, security and privacy issues for 6G remain largely in concept. This survey provides a systematic overview of security and privacy issues based on prospective technologies for 6G in the physical, connection, and service layers, as well as through lessons learned from the failures of existing security architectures and state-of-the-art defenses. Two key lessons learned are as follows. First, other than inheriting vulnerabilities from the previous generations, 6G has new threat vectors from new radio technologies, such as the exposed location of radio stripes in ultra-massive MIMO systems at Terahertz bands and attacks against pervasive intelligence. Second, physical layer protection, deep network slicing, quantum-safe communications, artificial intelligence (AI) security, platform-agnostic security, real-time adaptive security, and novel data protection mechanisms such as distributed ledgers and differential privacy are the top promising techniques to mitigate the attack magnitude and personal data breaches substantially.

... read more

Topics: Data Protection Act 1998 (56%), Data breach (53%)

1 Citations


Journal ArticleDOI: 10.1016/J.COSE.2021.102462
Abstract: This paper focuses on how zoom touchscreen gestures can be used to continuously authenticate and identify smartphone users. The zoom gesture is critically under-researched as a behavioral biometric despite richness of data found in this gesture. Furthermore, analysing how the zoom gesture performs over time is a novel line of inquiry. Zoom samples from three different data collection sessions were sourced. In these sessions, each participant zoomed in and out on three images. Eighty-five features were extracted from each gesture. The classification models used were Support Vector Machine (SVM), Random Forest (RF), and K-nearest Neighbor (KNN). The best authentication performance of AUC 0.937 and EER 10.6% were achieved using the SVM classifier. The best identification performance of 65.5% accuracy, 69.6% precision, and 67.9% recall were achieved using the RF classifier. In terms of stability over time, SVM proved to be the most stable classifier, with an AUC degradation of only 0.007 after two weeks had elapsed. This analysis proves that zoom gestures demonstrate promise for use in continuous smartphone authentication and identification applications.

... read more

Topics: Zoom (56%), Support vector machine (50%), Touchscreen (50%)

Book ChapterDOI: 10.1007/978-3-030-87007-2_33
13 Sep 2021-
Abstract: Biometric Wrist Authentication (BWA) is one of the best-known authentication schemes in many access control systems. The use of fingerprint biometrics as humans attempt to communicate with robots/machines, and their physical environments have inherent setbacks. However, various efforts have been proposed to fix the limitations. Most biometric efforts suffer from lack of computational derivatives and do not support optimal image compression. Motivated by these concerns, the goal of this paper is fivefold. First, we proposed BWA using Discrete Cosine Transform (DCT) to compress palm print images and develop Wrist-Print Biometric Identification System (WPBIS). Second, we developed a process model for DCT and characterized it for wrist templates considering both original and decoded images. Third, Bits per pixel (Bpp) and Compression ratio (Cr) for a wrist template/bioscript are used as metrics for evaluation. Fourth, after adopting various timestamps, we observed that the image template Bpp yielded 1.256 Bpp and compression of 63.26% based on DCT. Fifth, we showed a typical experimental scenario with a digital signal processor feeding images with DCT. Identification and verification of various wrist-prints (test-point samples) are equally carried out. From the results, WPBIS DCT offered higher image intensity compared with Wavelet transform.

... read more

Topics: Discrete cosine transform (58%), Biometrics (55%), Image compression (54%) ... show more

Book ChapterDOI: 10.1007/978-3-030-85893-3_9
23 Sep 2021-
Abstract: Modern technologies of authentication and authorization of access play a significant role in ensuring the protection of information in various practical applications. We consider the most convenient and used in modern mobile gadgets face authentication, ie when the primary information to provide access are certain features of biometric images of the user’s face. Most of the systems use intelligent processing of biometric images, in particular, artificial intelligence technology and deep learning. But at the same time, as always in cybersecurity, technologies for violating biometric authentication are being studied and researched. In particular, to date, the most common attack is substitution (spoofing), ie when attackers use pre-recorded biometric images to gain unauthorized access to critical information. For example, this could be a photo and/or video image of a person used to unlock their smartphone. Protection against such attacks is very difficult, because it involves the development and study of technologies for detecting signs of life. The most promising in this direction are artificial intelligence techniques, in particular, convolutional neural networks (CNN). This is the practical application of intelligent processing of biometric images and is studied in this article. We review various CNN settings and configurations and experimentally investigate their effect on the effectiveness of signs of life detection. For this purpose, success and failure indicators of the first and second kind are used, which are estimated by the values of cross entropy. These are reliable and reproducible indicators that characterize the effectiveness of protection against spoofing attacks on biometric authentication on the face. The world-famous TensorFlow and OpenCV libraries are used for field experiments, photos and videos of various users are used as source data, including Replay-Attack Database from Idiap Research Institute.

... read more

Topics: Spoofing attack (59%), Biometrics (54%), Authentication (53%)

Open accessJournal ArticleDOI: 10.1109/COMST.2021.3108618
Van-Linh Nguyen1, Po-Ching Lin1, Bo-Chao Cheng1, Ren-Hung Hwang1  +1 moreInstitutions (1)
Abstract: Sixth-generation (6G) mobile networks will have to cope with diverse threats on a space-air-ground integrated network environment, novel technologies, and an accessible user information explosion. However, for now, security and privacy issues for 6G remain largely in concept. This survey provides a systematic overview of security and privacy issues based on prospective technologies for 6G in the physical, connection, and service layers, as well as through lessons learned from the failures of existing security architectures and state-of-the-art defenses. Two key lessons learned are as follows. First, other than inheriting vulnerabilities from the previous generations, 6G has new threat vectors from new radio technologies, such as the exposed location of radio stripes in ultra-massive MIMO systems at Terahertz bands and attacks against pervasive intelligence. Second, physical layer protection, deep network slicing, quantum-safe communications, artificial intelligence (AI) security, platform-agnostic security, real-time adaptive security, and novel data protection mechanisms such as distributed ledgers and differential privacy are the top promising techniques to mitigate the attack magnitude and personal data breaches substantially.

... read more

Topics: Data Protection Act 1998 (56%), Data breach (53%)

References
  More

61 results found


Open accessProceedings Article
Diederik P. Kingma1, Jimmy Ba2Institutions (2)
01 Jan 2015-
Abstract: We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.

... read more

Topics: Stochastic optimization (63%), Convex optimization (54%), Rate of convergence (52%) ... show more

78,539 Citations


Proceedings ArticleDOI: 10.1109/CVPR.2014.220
23 Jun 2014-
Abstract: In modern face recognition, the conventional pipeline consists of four stages: detect => align => represent => classify. We revisit both the alignment step and the representation step by employing explicit 3D face modeling in order to apply a piecewise affine transformation, and derive a face representation from a nine-layer deep neural network. This deep network involves more than 120 million parameters using several locally connected layers without weight sharing, rather than the standard convolutional layers. Thus we trained it on the largest facial dataset to-date, an identity labeled dataset of four million facial images belonging to more than 4, 000 identities. The learned representations coupling the accurate model-based alignment with the large facial database generalize remarkably well to faces in unconstrained environments, even with a simple classifier. Our method reaches an accuracy of 97.35% on the Labeled Faces in the Wild (LFW) dataset, reducing the error of the current state of the art by more than 27%, closely approaching human-level performance.

... read more

5,249 Citations


Open accessProceedings ArticleDOI: 10.5244/C.29.41
01 Jan 2015-
Abstract: The goal of this paper is face recognition – from either a single photograph or from a set of faces tracked in a video. Recent progress in this area has been due to two factors: (i) end to end learning for the task using a convolutional neural network (CNN), and (ii) the availability of very large scale training datasets. We make two contributions: first, we show how a very large scale dataset (2.6M images, over 2.6K people) can be assembled by a combination of automation and human in the loop, and discuss the trade off between data purity and time; second, we traverse through the complexities of deep network training and face recognition to present methods and procedures to achieve comparable state of the art results on the standard LFW and YTF face benchmarks.

... read more

4,347 Citations


Proceedings ArticleDOI: 10.1109/CVPR.2006.100
Raia Hadsell1, Sumit Chopra1, Yann LeCun1Institutions (1)
17 Jun 2006-
Abstract: Dimensionality reduction involves mapping a set of high dimensional input points onto a low dimensional manifold so that 'similar" points in input space are mapped to nearby points on the manifold. We present a method - called Dimensionality Reduction by Learning an Invariant Mapping (DrLIM) - for learning a globally coherent nonlinear function that maps the data evenly to the output manifold. The learning relies solely on neighborhood relationships and does not require any distancemeasure in the input space. The method can learn mappings that are invariant to certain transformations of the inputs, as is demonstrated with a number of experiments. Comparisons are made to other techniques, in particular LLE.

... read more

Topics: Diffusion map (64%), Dimensionality reduction (62%), Invariant (mathematics) (58%) ... show more

3,011 Citations


Open accessProceedings ArticleDOI: 10.1109/CVPR.2017.713
Weiyang Liu1, Yandong Wen2, Zhiding Yu2, Ming Li3  +2 moreInstitutions (3)
26 Apr 2017-
Abstract: This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. However, few existing algorithms can effectively achieve this criterion. To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features. Geometrically, A-Softmax loss can be viewed as imposing discriminative constraints on a hypersphere manifold, which intrinsically matches the prior that faces also lie on a manifold. Moreover, the size of angular margin can be quantitatively adjusted by a parameter m. We further derive specific m to approximate the ideal feature criterion. Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge 1 show the superiority of A-Softmax loss in FR tasks.

... read more

Topics: Hypersphere (60%), Facial recognition system (52%), Convolutional neural network (51%) ... show more

1,398 Citations


Performance
Metrics
No. of citations received by the Paper in previous years
YearCitations
20216